Customer: Semiconductor Company
Location: Us-West1, Bay Area
Partner: Pluto7 Consulting Inc.
Customers Specific Challenges ‘pre-Google’ Solution:
Semiconductor Company Supply Chain is seeking to proactively reduce excess and obsolete inventory through better forecasting . The key drivers of these are as follows:
Currently business leads are hesitant to use automated forecasts for unique products since the accuracy of unique product forecast is low and cannot be relied on.
Semiconductor Company was also looking for better control over supply management, especially around usage multi-echelon inventory and inventory postponement techniques by leveraging forecast to determine what inventory to hold at lower levels that can be used to fulfil different demand while postponing the specific inventory creation in line with accurate forecasts thereby reducing probability of excess inventory.
Semiconductor Company is operating at Excess and Obsolete levels as 30% over Build Plan inventory.
GCP Components Used:
BiqQuery, Cloud SQL, Cloud Storage, Compute Engine, DataPrep Data Lab,Data Studio and Cloud ML.
Customer Acknowledgement of the ML work done on GCP:
“Thanks to you, Ram ( Pluto7), Tejas(Pluto7) and customer for getting us to a working POC model to get an understanding on how ML can be used for one aspect of business operations. I really appreciate the teams effort and having patiently explain me different models, transformation and other details during the detailed working sessions every week during this phase to get to a working model.” – Business Ops Leader at Semiconductor Company
Benefits Customer Has Seen so Far In Regards To Demand Forecasting:
The customer company is now aware of the capabilities of GCP and abilities of the ML model to forecast products better than traditional methods with improved accuracy. In some scenarios the ML model was better than human 8 out of 10 times. The customer was operating at excess and obsolete levels at high W.R.T the build plan inventory and it is expected to eventually improve once the forecasting ML model adoption evolves over rest of 2018.
for our use cases!